Performance assessment of GIS-based spatial clustering methods in forest fire data

Abstract Forest fires are a significant global issue, devastating large forest areas each year. Effective prevention and control are essential. Geographic Information System (GIS)-based spatial clustering methods are commonly used to manage forest fire risks. However, these methods rely on different mathematical foundations and parameters, resulting in varied hotspot maps. Consequently, areas identified as hotspots by one method may not be significant or may even be classified as cold spots by another. This study utilized forest fire data from 2021 and 2022 in Türkiye to conduct spatial clustering analyses using three methods: Getis Ord Gi*, Anselin Local Moran's I, and Kernel Density Estimation. The aim was to identify high-risk forest fire areas. The effectiveness of these methods was evaluated based on Hit Rate (HR), Predictive Accuracy Index (PAI), and Recapture Rate Index (RRI). The study concluded which method was most suitable for detecting risky forest fire areas in the region. This research fills a gap in the literature by providing a comparative performance evaluation of spatial clustering methods for forest fire risk assessment, offering valuable insights for future studies in this field.